Intelligent reconstruction and prediction system and method of three-dimensional wave field in real ocean environment

Abstract

The present invention discloses an intelligent reconstruction and prediction system and method of a three-dimensional wave field in a real ocean environment, which belongs to the field of marine engineering technology, and includes an unmanned aerial vehicle image capture module, an image preprocessing module, a feature recognition and parameter extraction module, a low-contrast and high-noise image processing module, a wave crest and trough recognition module, a wave characteristic analysis module, a wave characteristic analysis and prediction module, an unmanned aerial vehicle position adjustment module, and a wave simulation and evolution module. By integrating advanced image processing technology and artificial intelligence algorithms, the spatial wavefront resolution of free waves and high-order bound waves is improved, and accurate three-dimensional wave fields are synthesized by synchronous superposition. It is possible to use multiple high-altitude unmanned aerial vehicles in real time to collect wavefront images and perform wave feature recognition and parameter extraction under various sea conditions, and effectively realize the three-dimensional reconstruction of near-field, mid-field and far-field wavefronts in the ocean based on the wave evolution physical control equation.

Description

Technical Field

[0001] The invention relates to an intelligent reconstruction and prediction system and method of a three-dimensional wave field in a real ocean environment, belonging to the technical field of ocean engineering experimental devices. Background Art

[0002] Some progress has been made in the study of extreme waves encountered by ships and marine structures under high sea conditions in the prior art, but most of the research is limited to numerical simulation and physical tank experiments. The wave information used is completely based on statistical data and is limited to two-dimensional waves. The wave field generated after reconstruction is quite different from the real ocean environment. Patent publication number CN106990404A discloses an automatic calibration algorithm for inverting sea surface wave height using navigation X-band radar, which uses an automatic calibration algorithm for inverting real sea surface wave height using navigation X-band radar. However, this technology is only applicable to the wave time wave train of a single measuring point and cannot reconstruct spatial three-dimensional information. Another example is the patent publication number CN113970756A, which discloses a wave laser measurement device and a three-dimensional wave field spatiotemporal inversion and reconstruction method. It uses a laser radar measurement device to invert and reconstruct the three-dimensional wave field, but the device is located on the mast of the ship. When the radar scanning time is too long, the waves are constantly in motion and change, and the instantaneous spatial wave surface map cannot be accurately obtained. When the scanning time is too short, the measurement area is very limited and it is impossible to capture enough wave field information. In addition, in high sea conditions, this method is greatly affected by the movement of the ship, and it is difficult to complete the accurate measurement of the initial wave surface. The Chinese patent publication number CN110763189A discloses an experimental device and method for measuring sea wave height based on binocular vision. Its application scenario is limited to waves generated by laboratory wave-making plates. It has not been technically designed for complex three-dimensional waves in actual seas, and it is even more impossible to achieve the forecast analysis of far-field waves.

[0003] In summary, there is no technology that uses multiple drones to cross-acquire three-dimensional images in real time, nor has there been a study on extracting the characteristics of measured three-dimensional waves at sea through parallax analysis combined with deep learning algorithm analysis, and there is no method that uses data assimilation algorithm and physical control equations to predict the temporal and spatial evolution of three-dimensional waves in distant target areas based on the real-time spatial stereo wave field information of the measurement area. In actual engineering, in order to ensure the safety of ships and marine structures, very stringent requirements are usually made on the surrounding sea conditions during their operation. In addition, wave load forecasting and ship and marine structure structural reliability analysis based only on time domain statistical data will lead to great uncertainty in the results. In order to maximize the time and ability of ships and marine structures to operate in different sea conditions, it is urgent to develop an intelligent reconstruction and rapid prediction system and method of three-dimensional wave fields based on the fusion of deep learning and wave control equations. Summary of the invention

[0004] In view of the deficiencies of the prior art, the purpose of the present invention is to provide an intelligent reconstruction and prediction system and method for three-dimensional wave fields in a real ocean environment, which solves the problems encountered in the prior art.

[0005] The intelligent reconstruction and prediction method of three-dimensional wave field in a real ocean environment described in the present invention comprises the following steps:

[0006] S1: Use drones to capture wave image information of the sea area and transmit it to the wave image processing subsystem in real time via wireless signals;

[0007] S2: The wave image processing subsystem performs image preprocessing on the received wave image to enhance the wave imaging quality;

[0008] S3: Use an optimized stereo matching algorithm to extract disparity information from dual transient images, use deep learning technology to identify and process the three-dimensional wave surface, analyze and calculate the physical characteristics of the spatial wave field, and use artificial neural network algorithms to improve the feature recognition and parameter extraction accuracy of wave surface images;

[0009] S4: The training model is used to process wavefront images in low-contrast and high-noise environments, the visual system model is used to reconstruct the wave 3D point cloud image, and the moving least squares method is used to perform surface fitting on the discrete wavefront to achieve real-time reconstruction of the initial 3D wavefront in the monitoring area.

[0010] S5: The wave feature extraction subsystem analyzes the three-dimensional wave surface matrix data extracted based on the image, uses the discrete surface gradient operator to calculate the normal vector information of the wave surface, and combines the local peak value of the free liquid surface rise to realize the recognition and processing of the wave crest and trough features in the three-dimensional irregular sea conditions;

[0011] S6: Use the window function to process the boundary conditions of the sliding area, perform two-dimensional Fourier transform to obtain the energy spectrum, and estimate the main propagation direction of the wave, significant wave height, main wave wavelength and spectrum peak period;

[0012] S7: Use the composite neural network algorithm to realize the analysis and prediction of wave characteristics that interact and fuse in time and space dimensions, use the attention mechanism and regularization technology to adjust the weights, and establish an effective mapping relationship between the spatiotemporal information of the wave surface and the characteristic parameters of the wave;

[0013] S8: Based on the main wave direction information of the visual measurement area, the relative spatial position of the UAV and the seakeeping assessment area of ​​the marine structure is adjusted, the information of the wave feature extraction subsystem is updated in real time, and the feature parameters are re-evaluated;

[0014] S9: The wave evolution control subsystem uses an indicator model that incorporates directional diffusion to calculate the BFI index, screen the best wave control equation, use the random amplitude and phase wave model to deduce the wave change law, use the nonlinear wave equation to simulate the wave evolution process, use the Kalman filter algorithm to assimilate the extracted wave surface data, and simultaneously superimpose and synthesize an accurate three-dimensional wave field.

[0015] Furthermore, the image preprocessing in step S2 includes: enhancing the wave imaging quality by using an image preprocessing algorithm, including denoising, contrast adjustment and color correction.

[0016] Furthermore, in step S3, to address the problem of image degradation under different lighting conditions, a deep learning image enhancement algorithm that integrates RGB and HSV dual color spaces with multiple neural sub-networks is used for processing and analysis.

[0017] Furthermore, in step S7, the analysis and extraction of time domain wave information are realized synchronously. Based on the mapping relationship between the time variation sequence of wave surface information in the selected spatial grid points and wave characteristic parameters such as effective wave height and spectrum peak period, a partial convolution FasterNet model is introduced, and Conv2D and Block operations are performed with different numbers of convolution kernels, numbers of blocks and step sizes to realize the analysis and prediction of single-point wave characteristic parameters.

[0018] Furthermore, in S9, the wave surface of the random amplitude phase wave model η The equation is:

[0019] (1)

[0020] in, is the amplitude, is the wave number, For the direction of propagation, is the frequency, is the initial phase, t For time, and Indicates the position coordinates on the horizontal plane;

[0021] The nonlinear wave governing equation is

[0022] (2)

[0023] in: is the wave envelope function, represents the imaginary unit, represents the conjugate complex number, the velocity potential of the uniform flow induced by the wave The solution equation is:

[0024] (3)

[0025] (4)

[0026] (5)

[0027] Where: All the formulas (3), (4) and (5) about , , The subscripts of are the partial derivatives of the corresponding functions. Both time and space are described by dimensionless characteristic variables. is the second-order differential operator representing the conservation of mass, Represents water depth. Represents the three-dimensional vertical coordinates, and the free surface displacement can be reconstructed in multiple orders using the following formula:

[0028] (6)

[0029] in: represents the complex conjugate, represents the exponential function, the mean , second order Bound waves and third order The expression of bound wave is

[0030] (7)

[0031] (8)

[0032] (9)

[0033] in, is the wave envelope function along the space Seeking guidance, is the potential function of uniform flow along the space Derivative.

[0034] Furthermore, it also includes data assimilation and analysis steps, using the Kalman filter algorithm to assimilate the extracted wavefront data, using the Hilbert transform to obtain the complex plane information of the initial wave, combining the wave group velocity to convert the time and space characteristics, extracting the wavenumber spectrum information of the target prediction area, and performing linear interpolation through inverse Fourier transform to improve the spatial wavefront resolution.

[0035] The intelligent reconstruction and prediction system of three-dimensional wave fields in a real ocean environment described in the present invention comprises:

[0036] The UAV image capture module uses the UAV to synchronously capture the wave image information of a specific sea area and transmits it to the wave image intelligent processing subsystem in real time via wireless signals;

[0037] The image preprocessing module uses an image preprocessing algorithm to perform denoising, contrast adjustment and color correction on the received wave image to enhance the wave imaging quality;

[0038] The 3D wavefront reconstruction module uses an optimized stereo matching algorithm to extract disparity information from dual transient images, and uses deep learning technology to identify and process the 3D wavefront, analyzing and calculating the physical characteristics of the spatial wave field;

[0039] The feature recognition and parameter extraction module uses artificial neural network algorithms to improve the accuracy of feature recognition and parameter extraction of wavefront images;

[0040] The low-contrast and high-noise image processing module processes wave surface images in low-contrast and high-noise environments through a specially trained model, reconstructs the wave 3D point cloud using the visual system model, and uses the moving least squares method for surface fitting;

[0041] The wave crest and trough recognition module uses the discrete surface gradient operator to calculate the normal vector information of the wave surface based on the three-dimensional wave surface matrix data extracted from the image, and realizes the recognition and processing of wave crest and trough features in three-dimensional irregular sea conditions;

[0042] The wave characteristics analysis module uses the window function to process the boundary conditions of the sliding area, performs two-dimensional Fourier transform to obtain the energy spectrum, and estimates the main propagation direction of the wave, significant wave height, main wave wavelength and spectrum peak period;

[0043] The wave characteristic analysis and prediction module uses a composite neural network algorithm to realize the analysis and prediction of wave characteristics that are interactively integrated in time and space dimensions, and establish an effective mapping relationship between wave surface time and space information and wave characteristic parameters;

[0044] The UAV position adjustment module adjusts the relative spatial position of the UAV and the seakeeping assessment area of ​​the marine structure based on the main wave direction information of the visual measurement area, updates the information of the wave feature extraction subsystem in real time and re-evaluates the feature parameters;

[0045] The wave simulation and evolution module uses an indicator model that incorporates directional diffusion to calculate the BFI index, screens the optimal wave control equation, uses a random amplitude and phase wave model and an improved nonlinear wave equation to simulate the wave evolution process, and uses the Kalman filter algorithm to assimilate the extracted wave surface data to improve the spatial wave surface resolution.

[0046] Furthermore, it also includes a user interaction interface, which provides an interface for users to interact with the system, allowing users to input parameters, view analysis results and receive decision support information.

[0047] Furthermore, it also includes a data storage module for storing captured wave image information, processed image data, extracted wave characteristic parameters and comprehensive evaluation reports.

[0048] Furthermore, a remote monitoring module is included, which allows a remote user to view the system status, monitor the position of the drone and receive real-time wave characteristic information in real time through a network connection.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] 1. The present invention uses drone high-altitude camera technology to carry out stereoscopic reconstruction of real ocean three-dimensional wave surfaces, breaking through the limitations of traditional buoy and laser radar measurement technology. Drone high-altitude photography can not only avoid the interference caused by platform movement during measurement, but also effectively capture the trough information behind the large wave crest. The effective measurement area is large, the wave surface reconstruction is real-time, and its maneuverability is excellent, and the position of the measurement area can be adjusted in time. In addition, the principle of multiple stereo interactive parallax can improve spatial resolution and accuracy.

[0051] 2. This invention introduces deep learning algorithms into the measurement, reconstruction and real-time analysis of three-dimensional wavefronts for the first time, and uses improved convolutional neural networks to extract image features and perform high-precision stereo matching and point cloud post-processing. The deep learning model can achieve more accurate wavefront feature recognition and parallax calculation in complex ocean environments, greatly improving the accuracy of wavefront reconstruction. In particular, for low-contrast wavefront images, the deep learning algorithm demonstrates its strong robustness and can effectively deal with scenes that are difficult to handle with traditional analysis methods.

[0052] 3. The present invention combines the Kalman filter algorithm with the wave physics control equation for the first time, sequentially assimilates the extracted spatial three-dimensional wave surface information, obtains the optimal estimate of the current state by weighted averaging the high-altitude observation values ​​and the model prediction values, and selects the best control model based on the exponential factor. The relevant parameters of the linear dispersion and nonlinear focusing control items in the model are adjusted in real time in combination with the field intensity changes, ensuring the correct wave evolution mechanism while adapting to the complex and changeable marine environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is an overall logical flow chart of the intelligent reconstruction and rapid prediction method of three-dimensional wave field of the present invention;

[0054] Figure 2 A schematic diagram showing the relationship between the three subsystems involved in the present invention;

[0055] Figure 3 This is a schematic diagram of the spatial layout when the present invention actually works effectively;

[0056] Figure 4Statistical distribution diagrams of wave surfaces in the near field, midfield and far field of three-dimensional waves under high sea conditions obtained in Example 1 of the present invention; wherein, (a) is a statistical distribution diagram of wave surfaces in the near field of three-dimensional waves under high sea conditions; (b) is a statistical distribution diagram of wave surfaces in the midfield of three-dimensional waves under high sea conditions; (c) is a statistical distribution diagram of wave surfaces in the far field of three-dimensional waves under high sea conditions.

[0057] Figure 5 Statistical distribution diagrams of wave surfaces of near field, mid-field and far field of three-dimensional waves under moderate sea conditions obtained in Example 1 of the present invention; wherein (a) is a statistical distribution diagram of wave surfaces of near field of three-dimensional waves under moderate sea conditions; (b) is a statistical distribution diagram of wave surfaces of mid-field of three-dimensional waves under moderate sea conditions; (c) is a statistical distribution diagram of wave surfaces of far field of three-dimensional waves under moderate sea conditions;

[0058] Figure 6 The law of the evolution of the wave number spectrum of the three-dimensional wave field with time and space in Example 1 of the present invention; Figure (a) is the law of the evolution of the wave number spectrum of the near field of the three-dimensional wave field with time and space; Figure (b) is the law of the evolution of the wave number spectrum of the field in the three-dimensional wave field with time and space; Figure (c) is the law of the evolution of the wave number spectrum of the far field of the three-dimensional wave field with time and space;

[0059] Figure 7 The law of the energy spectrum of the three-dimensional wave field in Example 1 of the present invention evolving with time and space; (a) in the figure is the law of the energy spectrum of the near field of the three-dimensional wave field evolving with time and space; (b) in the figure is the law of the energy spectrum of the field in the three-dimensional wave field evolving with time and space; (c) in the figure is the law of the energy spectrum of the far field of the three-dimensional wave field evolving with time and space;

[0060] Figure 8 This is a wave surface display diagram of the three-dimensional wave field reconstructed at the midfield position according to Example 1 of the present invention. DETAILED DESCRIPTION

[0061] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0062] Embodiment 1:

[0063] like Figure 1 As shown, the intelligent reconstruction and prediction method of three-dimensional wave field in a real ocean environment described in the present invention comprises the following steps:

[0064] S1: Based on three high-altitude drones synchronously capturing wave image information of a specific sea area, wireless signals are used to transmit the information to the wave image intelligent processing subsystem in real time, and image preprocessing algorithms are used to enhance the wave imaging quality, realizing the functions of denoising, contrast adjustment and color correction respectively;

[0065] S2: Adopting the optimized stereo matching algorithm, the effective information of parallax can be accurately and efficiently extracted from the dual transient images, and the deep learning technology can be integrated to identify and process the three-dimensional wave surface, analyze and calculate the physical characteristics of the spatial wave field, and use the artificial neural network algorithm to improve the accuracy of feature recognition and parameter extraction of wave surface images.

[0066] S3: A specially trained model is used to process wavefront images in low-contrast and high-noise environments. The developed visual system model is used to reconstruct the wave 3D point cloud. The moving least squares method is used to fit the discrete wavefront to achieve real-time reconstruction of the initial 3D wavefront in the monitoring area.

[0067] S4: Analyze the three-dimensional wave surface matrix data extracted based on the image to effectively cut and divide its spatial area window, use the discrete surface gradient operator to calculate the normal vector information of the wave surface, and combine the local peak value of the free liquid surface rise to realize the recognition and processing of wave crests and troughs in three-dimensional irregular sea conditions.

[0068] S5: Use the window function to process the boundary conditions of the sliding area, carry out two-dimensional Fourier transform to obtain the energy spectrum, use the relationship between wave number and direction to estimate the main propagation direction of the wave, obtain the significant wave height by integrating the spectral area, analyze the main wave wavelength corresponding to the energy peak, and calculate the spectrum peak period using the wave dispersion relationship.

[0069] S6: Using a composite neural network algorithm composed of convolutions of different dimensions, the analysis and prediction of wave characteristics that interact and fuse in time and space dimensions are realized. The attention mechanism and regularization technology are used to adjust the weights, and the effective mapping relationship between the spatiotemporal information of the wave surface and the characteristic parameters of the wave is established through the big data regression training model.

[0070] S7: Based on the main wave direction information of the visual measurement area, adjust the relative spatial position of the UAV and the seakeeping capacity assessment area of ​​the marine structure to ensure that the effective measurement range of the stereoscopic vision is always located in the upstream direction of the incoming waves in the reconstructed far-field target area, update the information of the wave feature extraction subsystem in real time and re-evaluate the feature parameters.

[0071] S8: Use the index model that incorporates directional diffusion to calculate the BFI index and determine the physical field characteristics of the reconstructed sea area. Select the optimal wave control equation based on the index factor and the critical value, use the random amplitude and phase wave model to deduce the wave variation law dominated by the linear dispersion effect, and use the improved nonlinear wave equation to simulate the wave evolution process dominated by the interaction between high-order waves.

[0072] S9: The Kalman filter algorithm is used to sequentially assimilate the extracted wavefront data, the Hilbert transform is used to obtain the complex plane information of the initial wave, the time-space characteristics are converted in combination with the wave group velocity, the wave number spectrum information of the target prediction area is extracted according to the needs of the actual project, and linear interpolation is performed through the inverse Fourier transform to improve the spatial wavefront resolution of free waves and high-order bound waves, and a precise three-dimensional wave field is synthesized by synchronous superposition.

[0073] like Figure 2 As shown, the processing subsystem includes a wave image processing subsystem, a wave feature extraction subsystem and a wave evolution control subsystem. The input of the wave image processing subsystem determines the wave feature extraction subsystem, and the input of the wave feature extraction subsystem affects the wave evolution control subsystem.

[0074] In step S2, in order to solve the problem of image degradation under different lighting conditions, a deep learning image enhancement algorithm that integrates RGB and HSV dual color spaces with multiple neural sub-networks is used for processing and analysis. Deep learning is introduced into the stereo matching process, and an improved artificial intelligence algorithm is proposed to solve the problem that small parallax wavefront information is easily ignored and weak texture feature matching effect is poor during the reconstruction process.

[0075] In step S6, the analysis and extraction of time domain wave information are realized synchronously. Based on the mapping relationship between the time variation sequence of wave surface information in the selected spatial grid points and wave characteristic parameters such as effective wave height and spectrum peak period, the partial convolution FasterNet model is introduced, and Conv2D and Block operations are performed with different numbers of convolution kernels, numbers of blocks and step sizes, so as to more accurately realize the analysis and prediction of single-point wave characteristic parameters.

[0076] In step S8, the wavefront of the random amplitude phase wave model η The equation is:

[0077] (1)

[0078] in, is the amplitude, is the wave number, For the direction of propagation, is the frequency, is the initial phase, t For time, and Represents the position coordinates on the horizontal plane.

[0079] The improved nonlinear wave control equation is:

[0080] (2)

[0081] in: is the wave envelope function, represents the imaginary unit, represents the conjugate complex number, the velocity potential of the uniform flow induced by the wave The solution equation is:

[0082] (3)

[0083] (4)

[0084] (5)

[0085] All subscripts in the formula are partial derivatives of the corresponding functions, and both time and space are described by dimensionless characteristic variables. Represents water depth. Represents the three-dimensional vertical coordinate. The free surface displacement can be reconstructed in multiple orders using the following formula:

[0086] (6)

[0087] here represents the complex conjugate, represents the exponential function, the mean , second order Bound waves and third order The expression of bound wave is

[0088] (7)

[0089] (8)

[0090] (9)

[0091] The phase of the higher-order bound waves depends on the free waves, and the resulting uniform flow is closely related to the flow radiation stress.

[0092] This embodiment is mainly used to test and verify the accuracy of the two wave control models of the invention system in reconstructing and predicting the three-dimensional wave surface under different sea conditions. Figure 3 As shown, three drones are used. Figure 3 The spatial layout mode is used to place and install the measurement equipment. Figure 4 and Figure 5 The wave surface statistical regularity distribution diagrams of the three-dimensional wave near field, mid field and far field under high sea conditions and medium sea conditions obtained by the system designed by the present invention are respectively shown. Figure 4 (a) is the wave surface statistical distribution diagram of the three-dimensional wave near field under high sea conditions; Figure 4 Middle (b) is the wave surface statistical distribution diagram of the three-dimensional wave field under high sea conditions; Figure 4Middle (c) is the wave surface statistical distribution diagram of the three-dimensional wave far field under high sea conditions. Figure 5 (a) is the wave surface statistical distribution diagram of the three-dimensional wave near field under moderate sea conditions; Figure 5 (b) is the wave surface statistical distribution diagram of the three-dimensional wave field under moderate sea conditions; Figure 5 (c) is the wave surface statistical distribution diagram of the three-dimensional wave far field under moderate sea conditions.

[0093] Among them, the legend "Model 1" represents the wave surface change law predicted by the random amplitude phase model, and "Model 2" represents the wave surface evolution characteristics simulated by the improved nonlinear wave control equation. In order to more clearly describe the reconstruction accuracy of the two wave models, an indirect method is used to use the time domain wave data obtained by other measuring devices as a reference benchmark to compare and analyze the statistical laws of the time domain wave sequences extracted from the corresponding points of the two models of the system of the present invention. In addition, the theoretical distribution laws under the corresponding wave characteristic parameters are compared at the same time: Gaussian (linear) and Tayfun (second-order) distribution, so as to better reveal the physical evolution laws of three-dimensional waves in the reconstructed space field.

[0094] like Figure 4 As shown in the figure, in high sea conditions, the nonlinearity of waves gradually increases during the evolution process, and the height of the wave crest and the depth of the wave trough (the extreme values ​​at both ends of the figure) gradually increase, and the frequency of occurrence also gradually increases. At this time, neither the linear Gaussian distribution nor the second-order nonlinear Tayfun distribution can accurately describe the statistical distribution law of the wave surface. When the random amplitude phase model and the improved nonlinear wave model in the wave surface reconstruction system are used to predict and reconstruct the three-dimensional wave field, the statistical law of the overall wave surface lift height is relatively consistent, but with the increase of the propagation space distance, in terms of the amplitude prediction of larger wave crests and deeper wave troughs, both models will have certain differences from the experimental statistical reference values. This is because the physical assumptions of isotropy and weak stability at this time are difficult to be satisfied in high sea conditions. In general, the improved nonlinear wave model is obviously more accurate than the random amplitude phase model.

[0095] like Figure 5 As shown in the figure, in moderate sea conditions, due to the influence of directional diffusion, the nonlinear effect of three-dimensional waves gradually weakens as the propagation distance increases, and the three-dimensional wave surface distribution in the far field can be accurately described by the second-order nonlinear Tayfun distribution. At the same time, the amplitude difference of the predicted reconstruction results of the improved nonlinear wave model and the random amplitude phase model can be basically ignored, which is consistent with the overall statistical law of time domain waves measured at a specific location. It can be further inferred that the reconstruction accuracy of the system of the present invention will be higher in low sea conditions, and both models can effectively predict the change law of the three-dimensional wave surface in the farther field.

[0096] For a given specific sea area, the improved nonlinear wave model in the present invention is used to reconstruct the energy variation law of the spatial three-dimensional wave field in high sea conditions. The wave number spectrum generated by the prediction system reconstructed by the present invention is as follows: Figure 6 As shown, Figure 6 (a) shows the law of the evolution of the wave number spectrum of the near field of the three-dimensional wave field with time and space; Figure 6 (b) shows the law of the evolution of the wave number spectrum of the field in the three-dimensional wave field over time and space; Figure 6 (c) shows the law of the evolution of the wave number spectrum of the far field of the three-dimensional wave field with time and space; according to ( , ) and the wave propagation direction, it can be clearly determined that the main wave direction of the wave propagation is the defined global coordinate axis In the positive direction, as the propagation distance increases, the energy begins to be gradually transferred symmetrically to both sides of the main wave direction, which causes the overall nonlinearity of the wave to gradually weaken. The probability of higher crests and deeper troughs in the wave field is greatly reduced. The rules it presents are basically consistent with the theoretical analysis of directional diffusion, which once again proves the accuracy of the technology of the present invention in reconstructing and predicting three-dimensional wave fields.

[0097] In addition, the present invention can also accurately reconstruct and predict the time-space variation of the wave energy spectrum in the three-dimensional wave evolution process. As the propagation distance increases, Figure 7 As shown, the energy amplitude at the peak of the spectrum gradually decreases, and the reduced part is gradually transferred to the two ends of the spectrum, that is, the high frequency and low frequency, through the nonlinear interaction between waves. It is very obvious that the slope of the energy curve at the high frequency gradually stabilizes and presents a "saturation state", and the peak position of the spectrum "shifts downward". These reconstructed predictions of the physical change laws in the wave field are consistent with the results of the three-dimensional wave theory analysis, thus once again confirming the reliability of the technology of the present invention. As the width of the energy spectrum increases, the irregularity of the three-dimensional wave surface gradually increases.

[0098] like Figure 8 As shown, from the three-dimensional wave surface diagram reconstructed by the present invention at the midfield position, it can be seen that the wave surface of the entire wave field is extremely irregular. Although higher wave peaks still appear in some areas, the characteristics of its short bee waves are already very obvious, which is consistent with the real ocean environment.

[0099] Embodiment 2:

[0100] The intelligent reconstruction and prediction system of three-dimensional wave fields in a real ocean environment described in the present invention comprises:

[0101] The UAV image capture module uses the UAV to synchronously capture the wave image information of a specific sea area and transmits it to the wave image intelligent processing subsystem in real time via wireless signals;

[0102] The image preprocessing module uses an image preprocessing algorithm to perform denoising, contrast adjustment and color correction on the received wave image to enhance the wave imaging quality;

[0103] The 3D wavefront reconstruction module uses an optimized stereo matching algorithm to extract disparity information from dual transient images, and uses deep learning technology to identify and process the 3D wavefront, analyzing and calculating the physical characteristics of the spatial wave field;

[0104] The feature recognition and parameter extraction module uses artificial neural network algorithms to improve the accuracy of feature recognition and parameter extraction of wavefront images;

[0105] The low-contrast and high-noise image processing module processes wave surface images in low-contrast and high-noise environments through a specially trained model, reconstructs the wave 3D point cloud using the visual system model, and uses the moving least squares method for surface fitting;

[0106] The wave crest and trough recognition module uses the discrete surface gradient operator to calculate the normal vector information of the wave surface based on the three-dimensional wave surface matrix data extracted from the image, and realizes the recognition and processing of wave crest and trough features in three-dimensional irregular sea conditions;

[0107] The wave characteristics analysis module uses the window function to process the boundary conditions of the sliding area, performs two-dimensional Fourier transform to obtain the energy spectrum, and estimates the main propagation direction of the wave, significant wave height, main wave wavelength and spectrum peak period;

[0108] The wave characteristic analysis and prediction module uses a composite neural network algorithm to realize the analysis and prediction of wave characteristics that are interactively integrated in time and space dimensions, and establish an effective mapping relationship between wave surface time and space information and wave characteristic parameters;

[0109] The UAV position adjustment module adjusts the relative spatial position of the UAV and the seakeeping assessment area of ​​the marine structure based on the main wave direction information of the visual measurement area, updates the information of the wave feature extraction subsystem in real time and re-evaluates the feature parameters;

[0110] The wave simulation and evolution module uses an indicator model that incorporates directional diffusion to calculate the BFI index, screens the optimal wave control equation, uses a random amplitude and phase wave model and an improved nonlinear wave equation to simulate the wave evolution process, and uses the Kalman filter algorithm to assimilate the extracted wave surface data to improve the spatial wave surface resolution.

[0111] It also includes a user interaction interface, which provides an interface for users to interact with the system, allowing users to input parameters, view analysis results and receive decision support information.

[0112] It also includes a data storage module for storing captured wave image information, processed image data, extracted wave characteristic parameters and a comprehensive evaluation report.

[0113] Also included is a remote monitoring module that allows remote users to view system status, monitor drone location and receive real-time wave characteristics information in real time via a network connection.

[0114] The system of this embodiment uses drone high-altitude camera technology to provide information on the initial spatial three-dimensional wave field in the measurement area, improves the timeliness and accuracy of wave surface reconstruction by applying neural networks to improve image processing algorithms, effectively extracts spatiotemporal parameter information of the wave field based on a deep learning model, and uses a data assimilation algorithm combined with wave control equations to accurately and quickly predict the spatiotemporal evolution information of waves in the target area, so as to find the target of the best window period suitable for special operations at sea even in high sea conditions, and at the same time predict three-dimensional huge waves that may be encountered, and provide effective strategies to guide ships and marine structures to safely avoid abnormal waves.

[0115] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. An intelligent reconstruction and prediction method of three-dimensional wave field in a real ocean environment, characterized by: The following steps are involved: S1: Use drones to capture wave image information of the sea area and transmit it to the wave image processing subsystem in real time via wireless signals; S2: The wave image processing subsystem performs image preprocessing on the received wave image to enhance the wave imaging quality; S3: Use an optimized stereo matching algorithm to extract disparity information from dual transient images, use deep learning technology to identify and process the three-dimensional wave surface, analyze and calculate the physical characteristics of the spatial wave field, and use artificial neural network algorithms to improve the feature recognition and parameter extraction accuracy of wave surface images; S4: The training model is used to process wavefront images in low-contrast and high-noise environments, the visual system model is used to reconstruct the wave 3D point cloud image, and the moving least squares method is used to perform surface fitting on the discrete wavefront to achieve real-time reconstruction of the initial 3D wavefront in the monitoring area. S5: The wave feature extraction subsystem analyzes the three-dimensional wave surface matrix data extracted based on the image, uses the discrete surface gradient operator to calculate the normal vector information of the wave surface, and combines the local peak value of the free liquid surface rise to realize the recognition and processing of the wave crest and trough features in the three-dimensional irregular sea conditions; S6: Use the window function to process the boundary conditions of the sliding area, perform two-dimensional Fourier transform to obtain the energy spectrum, and estimate the main propagation direction of the wave, significant wave height, main wave wavelength and spectrum peak period; S7: Use the composite neural network algorithm to realize the analysis and prediction of wave characteristics that interact and fuse in time and space dimensions, use the attention mechanism and regularization technology to adjust the weights, and establish an effective mapping relationship between the spatiotemporal information of the wave surface and the characteristic parameters of the wave; S8: Based on the main wave direction information of the visual measurement area, the relative spatial position of the UAV and the seakeeping assessment area of ​​the marine structure is adjusted, the information of the wave feature extraction subsystem is updated in real time, and the feature parameters are re-evaluated; S9: The wave evolution control subsystem uses an index model that incorporates directional diffusion to calculate the BFI index, screen the best wave control equation, use the random amplitude and phase wave model to deduce the wave change law, use the nonlinear wave equation to simulate the wave evolution process, use the Kalman filter algorithm to assimilate the extracted wave surface data, and simultaneously superimpose and synthesize an accurate three-dimensional wave field; In step S9, the wave surface of the random amplitude phase wave model η The equation is: (1) in, is the amplitude, is the wave number, For the direction of propagation, is the frequency, is the initial phase, t For time, and Indicates the position coordinates on the horizontal plane; The nonlinear wave governing equation is (2) in: is the wave envelope function, represents the imaginary unit, represents the conjugate complex number, the velocity potential of the uniform flow induced by the wave The solution equation is: (3) (4) (5) Where: All the formulas (3), (4) and (5) about , , The subscripts of are the partial derivatives of the corresponding functions. Both time and space are described by dimensionless characteristic variables. is the second-order differential operator representing the conservation of mass, Represents water depth. Represents the three-dimensional vertical coordinates, and the free surface displacement can be reconstructed in multiple orders using the following formula: (6) in: represents the complex conjugate, represents the exponential function, the mean , second order Bound waves and third order The expression of bound wave is (7) (8) (9) in, is the wave envelope function along the space Seeking guidance, is the potential function of uniform flow along the space Derivative.

2. The intelligent reconstruction and prediction method of three-dimensional wave field in a real ocean environment according to claim 1 is characterized by: The image preprocessing in step S2 includes: using an image preprocessing algorithm to enhance the wave imaging quality, including denoising, contrast adjustment and color correction.

3. The intelligent reconstruction and prediction method of three-dimensional wave field in a real ocean environment according to claim 1 is characterized by: In the step S3, in order to solve the problem of image degradation under different lighting conditions, a deep learning image enhancement algorithm integrating RGB and HSV dual color spaces with multiple neural sub-networks is used for processing and analysis.

4. The intelligent reconstruction and prediction method of three-dimensional wave field in a real ocean environment according to claim 1 is characterized by: In the step S7, the analysis and extraction of time domain wave information are realized simultaneously. Based on the mapping relationship between the time variation sequence of wave surface information in the selected spatial grid point and the wave characteristic parameters of effective wave height and spectrum peak period, the partial convolution FasterNet model is introduced, and Conv2D and Block operations are performed using different numbers of convolution kernels, numbers of blocks and step sizes to realize the analysis and prediction of single-point wave characteristic parameters.

5. The intelligent reconstruction and prediction method of three-dimensional wave field in a real ocean environment according to claim 1 is characterized by: It also includes data assimilation and analysis steps. The Kalman filter algorithm is used to assimilate the extracted wave surface data, the Hilbert transform is used to obtain the complex plane information of the initial wave, the wave group velocity is combined to convert the time and space characteristics, the wave number spectrum information of the target prediction area is extracted, and linear interpolation is performed through inverse Fourier transform to improve the spatial wave surface resolution.

6. An intelligent reconstruction and prediction system for three-dimensional wave fields in a real ocean environment, based on the intelligent reconstruction and prediction method for three-dimensional wave fields in a real ocean environment according to any one of claims 1 to 5, characterized in that: The system comprises: The UAV image capture module uses the UAV to synchronously capture the wave image information of a specific sea area and transmits it to the wave image processing subsystem in real time via wireless signals; The image preprocessing module uses an image preprocessing algorithm to perform denoising, contrast adjustment and color correction on the received wave image to enhance the wave imaging quality; The 3D wavefront reconstruction module uses an optimized stereo matching algorithm to extract disparity information from dual transient images, and uses deep learning technology to identify and process the 3D wavefront, analyzing and calculating the physical characteristics of the spatial wave field; The feature recognition and parameter extraction module uses artificial neural network algorithms to improve the accuracy of feature recognition and parameter extraction of wavefront images; The low-contrast and high-noise image processing module processes wave surface images in low-contrast and high-noise environments by training models, reconstructs wave 3D point cloud images using the visual system model, and uses the moving least squares method for surface fitting; The wave crest and trough recognition module uses the discrete surface gradient operator to calculate the normal vector information of the wave surface based on the three-dimensional wave surface matrix data extracted from the image, and realizes the recognition and processing of wave crest and trough features in three-dimensional irregular sea conditions; The wave characteristics analysis module uses the window function to process the boundary conditions of the sliding area, performs two-dimensional Fourier transform to obtain the energy spectrum, and estimates the main propagation direction of the wave, significant wave height, main wave wavelength and spectrum peak period; The wave characteristic analysis and prediction module uses a composite neural network algorithm to realize the analysis and prediction of wave characteristics that are interactively integrated in time and space dimensions, and establish an effective mapping relationship between wave surface time and space information and wave characteristic parameters; The UAV position adjustment module adjusts the relative spatial position of the UAV and the seakeeping assessment area of ​​the marine structure based on the main wave direction information of the visual measurement area, updates the information of the wave feature extraction subsystem in real time and re-evaluates the feature parameters; The wave simulation and evolution module uses an indicator model that incorporates directional diffusion to calculate the BFI index, screens the optimal wave control equation, uses a random amplitude and phase wave model and an improved nonlinear wave equation to simulate the wave evolution process, and uses the Kalman filter algorithm to assimilate the extracted wave surface data to improve the spatial wave surface resolution.

7. The intelligent reconstruction and prediction system of three-dimensional wave fields in a real ocean environment according to claim 6, characterized in that: It also includes a user interaction interface, which provides an interface for users to interact with the system, allowing users to input parameters, view analysis results and receive decision support information.

8. The intelligent reconstruction and prediction system of three-dimensional wave fields in a real ocean environment according to claim 6, characterized in that: It also includes a data storage module for storing captured wave image information, processed image data, extracted wave characteristic parameters and a comprehensive evaluation report.

9. The intelligent reconstruction and prediction system of three-dimensional wave fields in a real ocean environment according to claim 6, characterized in that: Also included is a remote monitoring module that allows remote users to view system status, monitor drone location and receive real-time wave characteristics information in real time via a network connection.

Images